Measuring Life Expectancy and Lifespan Variation Inequalities at Older Ages

Are Ageing Societies Ageing Equally? Measuring Life Expectancy and Lifespan Variation Inequalities at Older Ages

Seaman, Rosie1* and Nepomuceno, Marília R.1

1Max Planck Institute for Demographic Research, Rostock, Germany

Abstract Socioeconomic inequalities in average life expectancy are considered to be the strongest reflection of inequalities in health and life chances but the WHO recommend that the distribution of health across all individuals also be monitored. This prompted studies into variation in age at death. Higher variation in death reflects greater heterogeneity or greater inequality. International evidence shows that the most deprived groups experience a double burden of mortality inequality: lower life expectancy and higher variation. At the national level, older ages have shown increasing variation in age at death. It is not clear if this national level trend is consistent across all socioeconomic groups. Measuring life expectancy and variation at older ages is important for understanding the extent to which ageing societies are ageing equally. We investigate life expectancy and lifespan variation changes between 1981 and 2011, at age 50, age 65 and age 75, using area-level deprivation to stratify the whole population of Scotland.

Keywords: heterogeneity in age at death, lifespan variation gap, socioeconomic inequality, inequality and ageing

Acknowledgements and funding This work was supported by the European Research Council (grant number 716323). We are grateful to the Census and vital statistics customer service teams at National Records of Scotland for providing data.

* Corresponding author: [email protected] 1 Background

2 Socioeconomic inequalities in life expectancy have been documented in every developed country 3 where measurement has taken place. A gradient in life expectancy is found whether education, 4 income, or occupation are used as a proxy measure for socioeconomic inequality with trend studies 5 indicating that the life expectancy gap between the most and least advantaged groups has widened 6 in developed countries (Auger, Harper and Barry 2013; Martikainen, Valkonen and Martelin 2001; 7 Tarkiainen et al. 2012). Although inequalities in life expectancy are considered to be the strongest 8 reflection of inequalities in health and life chances, recent attention has been paid to variation in age 9 at death.

10 Higher variation in death reflects greater heterogeneity or greater inequality. It is also in line with 11 the WHO recommendation to measure the distribution of health across all individuals (World Health 12 Organistation 2000). National variation trends have been investigated for many countries (Seaman, 13 Leyland and Popham 2016; Vaupel, Zhang and van Raalte 2011). In general, the desired correlation 14 between increasing life expectancy and decreasing variation in age at death has been achieved by 15 most countries. This is largely due to falling premature mortality rates historically outpacing falling 16 mortality rates at older ages (Vaupel et al. 2011).

17 In contrast to national trends, socioeconomic trends within countries are not consistently following 18 the desired correlation: the most deprived groups experience increasing life expectancy alongside 19 stagnating or even increasing variation in age at death (Brønnum-Hansen 2017; van Raalte, 20 Martikainen and Myrskylä 2014). Lower life expectancy and higher variation in age at death for the 21 most deprived socioeconomic groups can be considered as a double burden of mortality inequality.

22 While increases in life expectancy come from reducing mortality rates at any age, variation in age at 23 death decreases if there is a balance between avoiding premature deaths and avoiding older age 24 deaths. The distribution compresses if premature mortality rates are decreased over time. The 25 distribution expands when deaths at older ages are delayed (Vaupel et al. 2011). Therefore, 26 increases in life expectancy could be expected to continue with increased survival at older ages but 27 decreases in variation may be more uncertain. Increasing variation at older ages can on one hand be 28 considered a consequence of population ageing. On the other hand, if increases at older ages are 29 patterned by socioeconomic deprivation it may indicate that ageing is becoming an increasingly 30 unequal process. Regardless, measuring variation has implications in terms of future planning, 31 pensions and health care (van Raalte et al. 2014).

32 When looking at variation conditional upon survival to age 75 for national populations (Engelman, 33 Canudas-Romo and Agree 2010) found the level to be stable with some small increases since 1950. 34 Several studies have also documented socioeconomic inequalities (Sasson 2016; van Raalte et al. 35 2012; van Raalte et al. 2014) but these studies paid limited attention to inequalities at older ages. 36 We consider the life expectancy gap and the lifespan variation gap conditional upon survival to age 37 50, age 65 and age 75 between 1981 and 2011. We use population and mortality data that covers 38 the entire population of Scotland (c.a. 5 million) stratified by a validated measure of area-level 39 deprivation(Carstairs and Morris 1989). The next steps of this research will involve age-and cause 40 specific decomposition.

42 Data and methods

43 We use individual level mortality data and area-level population estimates that were obtained via a 44 commissioned request from National Records of Scotland (National Records of Scotland), alongside 45 the Carstairs score of area-level deprivation that is freely available to access online (Brown et al. 46 2014). After matching, these data were used to construct deprivation specific period life tables for 47 each census year, for males and females separately.

48 Population data are for each part-postcode sector in Scotland at each census year between 1981 and 49 2011, ensuring the most robust population estimates available that include the whole population. 50 There are approximately 1,010 part-postcode sectors in Scotland, with an average population size of 51 5,000. Population estimates obtained were stratified by single year of age and sex.

52 Mortality data include all individual deaths that occurred in Scotland in the years centred around 53 each census year. Each individual death record contains geographical information on place of usual 54 residence that we used to match to population estimates and the deprivation score data.

55 Traditional stratification measures such as income, occupation or education are often only available 56 for certain ages of the population meaning age truncation is required. Area-level measures of 57 deprivation are a unique alternative that overcome this limitation as they are applicable to the 58 entire population and are derived from routinely collective, administrative data (Kearns, Gibb and 59 Mackay 2000). In this study we use the Carstairs score, a standardised z-score derived from four 60 variables (overcrowding, male unemployment, no car ownership and low social class) that was 61 constructed specifically for measuring health inequalities (Carstairs and Morris 1989).

62 Smoothing and extrapolation

63 Census population estimates obtained were available up to different open-ended age intervals; 1981 64 was 85+, 1991 was 90+, 2001 was 85+ and 2011 was 95+. According to the 2011 Scottish female 65 period life table, over 45 percent of women survived to ages older than 85 (Human Mortality 66 Database 2018) . Given the likely differences in survivorship by deprivation quintile, using an open- 67 ended age interval of 85+ risked introducing biases in lifespan variation according to the proportions 68 surviving to age 85. Although a mortality rate for the entire population of Scotland was available 69 from the Human Mortality Database (HMD) up to an open-ended age interval of 110+ this could not 70 directly be used to derive any information on socioeconomic deprivation. Instead we used the HMD 71 smoothing method, a Kannisto logistic model, to extrapolate mortality rates to age 110+ for each 72 year, sex, and quintile separately. We report results using an open age category of 110+. 73 Specifically, we apply equations 64 and 65 from the HMD Methods Protocol version 6 (Wilmoth et al. 74 2017), but modified to use information from ages 75+ rather than 80+.

75 Measures of lifespan variation

† 76 We measure inequality in age at death using 푒푥 . This is the weighted average of remaining life 77 expectancy at each age, weighted by the number of lifetable deaths at each age. It has an intuitive 78 public health interpretation: the average number of years remaining at death and measures life 79 years lost when a death occurs (Vaupel and Canudas Romo 2003). The high correlation between 80 alternative indices of lifespan inequality suggests that the overall conclusions would not have

81 changed had an alternative measure been used (Németh 2017; van Raalte and Caswell 2013; Vaupel 82 et al. 2011; Wilmoth and Horiuchi 1999).

83 Preliminary descriptive results

84 Figure 1 shows remaining life expectancy for ages >50 to 110+ plotted against disparity in age at 85 death for ages>50 to 110+. It compares the most and least deprived quintile in 1981 and 2011. Three 86 ages are highlighted: 50, 65 and 75.

88 Figure 1 Relationship between remaining life expectancy and variation in age at death conditional 89 upon survival to ages >50, for most and least deprived quintile.

90 The overall patterns for males and females are similar but females have slightly higher remaining life 91 expectancy and slightly lower variation in age at death. The most striking finding in this figure is the 92 change between 1981 and 2011 for the least deprived quintile that is not present for the most 93 deprived quintile. Between 1981 and 2011, we see that remaining life expectancy for the least 94 deprived quintile has increased more than for the most deprived. Variation has decreased for those 95 age 50 from the least deprived, stayed constant for those age 65 and has slightly increased for those 96 age 75. This generally indicates increased longevity and decreasing inequality between individuals 97 for the least deprived. This is in contrast to the least deprived where variation in 2011 is consistently 98 higher across all ages than it was in 1981.

100

101 Discussion

102 Our initial results show that that ageing is an unequal process between socioeconomic groups and 103 within socioeconomic groups. Life expectancy has increased for older ages across all socioeconomic 104 groups but within in the context of higher absolute life expectancy and greater relative gains for the 105 least deprived. Variation does not show the same systematic gains. Variation in age at death has 106 increased at older ages for the most deprived but stayed fairly constant for the least deprived. This 107 means that older individuals from the most deprived quintile experience increasing uncertainty in 108 age at death. This finding builds upon an earlier study which showed a small expansion of life span 109 inequalities among older ages at the national level (Engelman et al. 2010) but we demonstrate that 110 this expansion of inequalities to be greater in magnitude for the most deprived in Scotland.

111 Variation partly reflects the fact that survival improvements at older ages have risen (Rau et al. 2008; 112 Wilmoth et al. 2000). Whether improved survival at extreme old age is positive or negative attribute 113 of population health depends on whether additional life years are lived in good or bad health. 114 Engelman et al. (2010) argue that the mortality composition of a population may be changed and 115 health disparities delayed to older ages if increased survival rates result in a more diverse population 116 and an increasing number of frail members of the population. Societies need to meet the needs of 117 an ageing population and measuring variation inequalities can help to inform social policies. For 118 example, increasing variation has implications at the individual and at the societal level in terms of 119 increased uncertainty surrounding pension planning and healthcare needs (van Raalte et al. 2014). 120 Our study indicates that increased uncertainty is burdened by the most vulnerable older individuals 121 who may be at greater risk of pension poverty. This is a particular concern in Scotland where 122 pensioner poverty is higher than the UK average (Social Meterics Commission 2018). Increasing 123 variation in age at death has implications for estimating the level of resources an ageing population 124 will likely need and reveals inequalities that are hidden behind average life expectancy.

125 Research outlook

126 The next steps of this project will decompose life expectancy and lifespan variation into the age and 127 cause specific contributions using a method of continuous change (Horiuchi, Wilmoth and Pletcher 128 2008). We will carry out the following decompositions (i) the change in life expectancy for the most 129 and for the least deprived quintile (ii) the change in lifespan variation for the most and for the least 130 deprived quintile (iii) the life expectancy gap between the most and the least deprived quintile and 131 (iv) the lifespan variation gap between the most and least deprived quintile. This will help to better 132 understand the reasons why we find life expectancy and lifespan variation inequalities at older ages.

133 References

134 Auger, N., S. Harper, and A.D. Barry. 2013. "Diverging socioeconomic inequality in life expectancy of 135 Francophones and Anglophones in Montréal, Québec: tobacco to blame?" Journal of Public Health 136 21(4):317-324. 137 Brønnum-Hansen, H. 2017. "Socially disparate trends in lifespan variation: a trend study on income 138 and mortality based on nationwide Danish register data." BMJ Open 7(5). 139 Brown, D., M. Allik, R. Dundas, and A.H. Leyland. 2014. "Carstairs Scores for Scottish Postcode 140 Sectors, Datazones and Output Areas from the 2011 Census." 141 Carstairs, V.and R. Morris. 1989. "Deprivation and mortality: an alternative to social class?" Journal 142 of Public Health 11(3):210-219.

143 Engelman, M., V. Canudas-Romo, and E.M. Agree. 2010. "The Implications of Increased Survivorship 144 for Mortality Variation in Aging Populations." Population and Development Review 36(3):511-539. 145 Horiuchi, S., J.R. Wilmoth, and S.D. Pletcher. 2008. "A decomposition method based on a model of 146 continuous change." Demography 45(4):785-801. 147 Human Mortality Database. 2018. University of California Berkeley (USA) and Max Planck Institute 148 for Demographic Research (Germany). 149 Kearns, A., K. Gibb, and D. Mackay. 2000. "Area deprivation in Scotland: A new assessment." Urban 150 studies 37(9):1535-1559. 151 Martikainen, P., T. Valkonen, and T. Martelin. 2001. "Change in male and female life expectancy by 152 social class: decomposition by age and cause of death in Finland 1971–95." Journal of epidemiology 153 and community health 55(7):494-499. 154 National Records of Scotland. "http://www.nrscotland.gov.uk [Data obtained: 2017]." 155 Németh, L. 2017. "Life expectancy versus lifespan inequality: A smudge or a clear relationship?" PLoS 156 ONE 12(9):e0185702. 157 Rau, R., E. Soroko, D. Jasilionis, and J.W. Vaupel. 2008. "Continued Reductions in Mortality at 158 Advanced Ages." Population and Development Review 34(4):747-768. 159 Sasson, I. 2016. "Trends in Life Expectancy and Lifespan Variation by Educational Attainment: United 160 States, 1990–2010." Demography:1-25. 161 Seaman, R., A.H. Leyland, and F. Popham. 2016. "How have trends in lifespan variation changed 162 since 1950? A comparative study of 17 Western European countries." The European Journal of Public 163 Health 26(2):360-362. 164 Social Meterics Commission. 2018. "A new measure of poverty for the UK." Legathum Institute. 165 Tarkiainen, L., P. Martikainen, M. Laaksonen, and T. Valkonen. 2012. "Trends in life expectancy by 166 income from 1988 to 2007: decomposition by age and cause of death." Journal of epidemiology and 167 community health 66(7):573-578. 168 van Raalte, A., A.E. Kunst, O. Lundberg, M. Leinsalu, P. Martikainen, B. Artnik, P. Deboosere, I. Stirbu, 169 B. Wojtyniak, and J.P. Mackenbach. 2012. "The contribution of educational inequalities to lifespan 170 variation." Popul Health Metr 10(1):3-3. 171 van Raalte, A.A.and H. Caswell. 2013. "Perturbation Analysis of Indices of Lifespan Variability." 172 Demography 50(5):1615-1640. 173 van Raalte, A.A., P. Martikainen, and M. Myrskylä. 2014. "Lifespan variation by occupational class: 174 compression or stagnation over time?" Demography 51(1):73-95. 175 Vaupel, J.W.and V. Canudas Romo. 2003. "Decomposing change in life expectancy: A bouquet of 176 formulas in honor of Nathan Keyfitz’s 90th birthday." Demography 40(2):201-216. 177 Vaupel, J.W., Z. Zhang, and A.A. van Raalte. 2011. "Life expectancy and disparity: an international 178 comparison of life table data." BMJ Open 1(1). 179 Wilmoth, J.R., K. Andreev, D. Jdanov, D.A. Glei, T. Riffe, C. Boe, M. Bubenheim, D. Philipov, V. 180 Shkolnikov, P. Vachon, C. Winant, and M. Barbiere. 2017. "Methods protocol for the human 181 mortality database." University of California, Berkeley, and Max Planck Institute for Demographic 182 Research, Rostock. URL: http://mortality. org [version 27/11/2017]. 183 Wilmoth, J.R., L.J. Deegan, H. Lundström, and S. Horiuchi. 2000. "Increase of Maximum Life-Span in 184 Sweden, 1861-1999." Science 289(5488):2366-2368. 185 Wilmoth, J.R.and S. Horiuchi. 1999. "Rectangularization revisited: Variability of age at death within 186 human populations." Demography 36(4):475-495. 187 World Health Organistation. 2000. "The world health report 2000: health systems: improving 188 performance." edited by W.H. Organization.